Method infer

samples/python/efficientnet/infer.py:99–124

Execute inference on a batch of images. The images should already be batched and preprocessed, as prepared by the ImageBatcher class. Memory copying to and from the GPU device will be performed here. :param batch: A numpy array holding the image batch. :param top: Th

(self, batch, top=1)

        return self.outputs[0]["shape"], self.outputs[0]["dtype"]

    def infer(self, batch, top=1):
        """
        Execute inference on a batch of images. The images should already be batched and preprocessed, as prepared by
        the ImageBatcher class. Memory copying to and from the GPU device will be performed here.
        :param batch: A numpy array holding the image batch.
        :param top: The number of classes to return as top_predicitons, in descending order by their score. By default,
        setting to one will return the same as the maximum score class. Useful for Top-5 accuracy metrics in validation.
        :return: Three items, as numpy arrays for each batch image: The maximum score class, the corresponding maximum
        score, and a list of the top N classes and scores.
        """
        # Prepare the output data
        output = np.zeros(*self.output_spec())

        # Process I/O and execute the network
        common.memcpy_host_to_device(self.inputs[0]["allocation"], np.ascontiguousarray(batch))
        self.context.execute_v2(self.allocations)
        common.memcpy_device_to_host(output, self.outputs[0]["allocation"])

        # Process the results
        classes = np.argmax(output, axis=1)
        scores = np.max(output, axis=1)
        top = min(top, output.shape[1])
        top_classes = np.flip(np.argsort(output, axis=1), axis=1)[:, 0:top]
        top_scores = np.flip(np.sort(output, axis=1), axis=1)[:, 0:top]

        return classes, scores, [top_classes, top_scores]


def main(args):